Children with specific language impairment (SLI) and developmental delays (DD) are in special need of ways to facilitate their grammatical development. If we could help these children elicit grammar-facilitating adult input (i.e., grammatical recasts), we would empower children to enhance their own grammatical development. One way to elicit grammatical recasts may be to facilitate more frequent child use of the types of utterances that mothers naturally recast. Mothers most frequently naturally recast children's multi-work and fully intelligible utterances. Therefore, an intervention that facilitates children's frequent and generalized use of multi-word, fully intelligible utterances may result in children receiving facilitating linguistic input in many more times, places, conversational contexts, and people that occurs in traditional language intervention. We plan to test five hypotheses. First, we predict that conversational language intervention will increase the children's intelligibility and length of utterance. Second, we predict that facilitating intelligibility and length of utterance will lead to the elicitation of more naturally-occurring maternal recasts. However, because no treatment is equally effective for all children, we also predict the group difference in naturally occurring maternal recasts will tend to be more pronounced in mothers with relatively high family resources. Third, we predict that the combination of the conversational language intervention and the subsequent increase in naturally-occurring maternal grammatical recasts will facilitate grammatical semantic relations and early syntactic and morphological development several months after the end of the intervention. Fourth, in addition to this main effect, we will examine the possibility that children with non-verbal intelligence in the normal range (i.e., SDL) will benefit faster from the conversational language intervention and subsequent increase in naturally-occurring maternal recasts that will children with general developmental delays (i.e., DD). Fifth, we predict that statistically controlling post- treatment naturally-occurring maternal recasting will render the treatment effect on later grammatical level non-significant. If it does, then there will be evidence that the degree to which mothers naturally recast after the treatment is responsible for, at least part of, the treatment effect on grammatical development at follow-ups.